AI Paper Atlas

Research map / Point clouds / Point cloud classification

Upright-Net: Learning Upright Orientation for 3D Point Cloud

Xufang Pang, Feng Li, Ning Ding, Xiaopin Zhong

CVPR 2022 · Point clouds › Point cloud classification · approach: Local feature aggregation networks · 11 citations (OpenAlex)

Check your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Estimate the natural base by point segmentation, then derive orientation from its fitted planeUpright-Net · UprightNet15 testAccuracy (%): 69.01 -> 91.80Improved
Estimate the natural base by point segmentation, then derive orientation from its fitted planeUpright-Net · UprightNet5 meanAccuracy: 28.40 -> 81.39Improved
Use Upright-Net for classification on rotationally augmented dataRobustPointSet · Rot/RotClassification accuracy: 58.95 -> 67.83Improved
Pre-train the Upright-Net encoder before downstream classificationRobustPointSet · Org/RotClassification accuracy: 14.18 -> 49.92Improved
Pre-train the Upright-Net encoder before downstream classificationRobustPointSet · Rot/RotClassification accuracy: 67.83 -> 79.62Improved
Replace the proposed segmentation backbone with PointNet (ablation)Upright-Net · UprightNet15 testAccuracy (%): 91.80 -> 77.68Hurt
Replace the proposed segmentation backbone with DGCNN (ablation)Upright-Net · UprightNet15 testAccuracy (%): 91.80 -> 90.26Hurt
Add fitting-residual and stability loss terms rather than using BCE alone (ablation)Upright-Net · UprightNet15 testAccuracy (%): 85.13 -> 91.80Improved
Use α1=1 and α2=0.1 instead of the reported default α1=0.1 and α2=1 (ablation)Upright-Net · UprightNet15 testAccuracy (%): 91.80 -> 92.47Improved
Use α1=1 and α2=1 instead of the reported default α1=0.1 and α2=1 (ablation)Upright-Net · UprightNet15 testAccuracy (%): 91.80 -> 91.22Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy